Method enqueue_batch

cactus/kernel/kernel_utils.h:359–377

        template<typename F>
        void enqueue_batch(size_t total_work, F task_func) {
            if (total_work == 0) return;

            const size_t num_tasks = std::min(num_workers_, total_work);
            const size_t per_worker = total_work / num_tasks;
            const size_t remainder = total_work % num_tasks;

            {
                std::lock_guard<std::mutex> lock(mutex);
                pending_tasks.fetch_add(num_tasks, std::memory_order_relaxed);

                for (size_t w = 0; w < num_tasks; ++w) {
                    size_t start = w * per_worker + std::min(w, remainder);
                    size_t end = start + per_worker + (w < remainder ? 1 : 0);
                    tasks.emplace_back([=]() { task_func(start, end); });
                }
            }
            work_available.notify_all();
        }

        void wait_all() {
            std::unique_lock<std::mutex> lock(mutex);